pugscode.org <<Overview Journal>> | pugs.kwiki.org | logged: xrl.us/e98m | FBSD_4.11:r1257(2/3212) win2k:r1197(105/3116) Linux:r1197(105/3116) MacOSX:r1197(105/3116) **[793 tests TODO]**
Set by kungfuftr on 27 March 2005.
Juerd Seeing autrijus' photo was rather disappointing - he doesn't look like the mad genius I had pictured... :) 00:27
He actually looks normal.
theorbtwo Mmmpf, I'm drifting something horrible. 00:29
metaperl Juerd, I feel sorry for you 00:50
PLP is nicer than Mason, yet they are madly popular... how does that make you feel? 00:51
Juerd "nicer"? 00:53
It's leight weight and fast
But Mason is a programming framework
They don't really compete.
s/leight/light/
And I should note that I don't use PLP much myself, except as a CGI-replacement 00:54
That is: "<:" has taken over "#!/usr/bin/perl"
Because this lets me develop as easy as things go with CGI, but with mod_perl's speed 00:55
metaperl I see... I am a CGI::Prototype fan 01:06
Juerd You can probably very well use that from within PLP 01:10
metaperl perhaps, but if I invented Seamstress, I need to use it 01:13
Juerd Why that? 01:14
xmath and I invented Attribute::Prototype, but you don't see any one of us use it :)
metaperl yes, I like the Seamstress approach though... it's a bit indirect, but it gives me the non-invasive approach that I demand 01:15
you know, you should ask to be in Perrin's "choosing a templating system" doc at perl.apache.org 01:16
Juerd Why? 01:25
metaperl exposure 01:27
Juerd Exposure for PLP isn't necessarily a good thing. 01:28
metaperl oh? why?
Juerd It needs to be rewritten and I don't want to spend the little time I have on that
metaperl oh 01:29
Juerd PLP works, but it's UGLY
And has some weird features that the world really would be better without.
theorb Sigh... I know I still have problems with testgraph, but they seem to be firefox bugs, and I haven't a clue what to do with them... anybody have another project to suggest? 02:14
stevan theorb: how about sleep? you sound tired :) 02:21
theorb I'm not tired, just depressed. 02:22
stevan nothingmuch and I were talking about forking the testgraph stuff off into its own module
it would be useful in other contexts 02:23
theorb nods.
I certianly wouldn't be opposed to such a thing. 02:24
jabbot pugs - 1264 - adding in two new tests, one for rx:perl 02:27
pugs - 1263 - making the exceptions lexical classes
stevan theorb: have you ever tried 'make optimized'? 02:28
theorb I have, but not for quite a while. 02:29
stevan I am wondering if it worth it?
I suppose I should give it a try
theorb In fact, come to think of it, I never have, but I'm the one who poked people into making it.
stevan :) 02:30
theorb Last I timed it, a compile-from-clean + run all tests ran longer with a -O1 and -O3 then with the default.
stevan but was that all in the compile time? 02:31
if you are looking for another project though, maybe you can port File::basename 02:32
iblech started it
but it doesn't work, he just ported
you could hack it like I did File::Spec
its frustrating and satisfiying at the same time 02:33
jabbot pugs - 1265 - un-TODO-ing the tests and making things 02:37
cm you guys are so motivated :) 02:41
stevan cm: it keeps me out of trouble :) 02:51
theorb It just keeps me depressed. 03:01
cm :) 03:06
does larry wall sometimes visit this channel? 03:09
theorb Not that I've seen.
kungfuftr larry, irc? 03:10
theorb He's been known to talk on occasion on the CB of Perlmonks. 03:11
cm what's "CB"? 03:12
theorb Chatterbox. Go to perlmonks.org, it's the thing near the top-left with conversation in it.
(Assuming anybody has said anything recently.)
cm ah ok
theorb Perlmonks: the reason I can put "god" on my resume and mean it. 03:14
kungfuftr anyone ever built maxdb at all? 03:22
jabbot pugs - 1266 - * fix warning 03:27
pugs - 1267 - fix pugsrun tests. '&>' will go backgoun 03:47
pugs - 1268 - r1269@scwnb: scw | 2005-03-27 21:45:10 03:57
pugs - 1269 - r1306@scwnb: scw | 2005-03-28 12:00:12 04:07
mugwump scw: use svk push --verbatim 04:10
scw mugwump: Err, there are two things but I cannat find a way to push twice 04:18
So I decided to push without verbatim.. 04:19
oop, I forgot to use --verbatim in r1269, is there some where to write it in config file? 04:21
jabbot pugs - 1270 - more work on Pod::Stream::Parser; Added 04:27
pugs - 1272 - Fix :key(val) and unTODO tests. 04:47
pugs - 1271 - Local redo.
gaal mornin 04:54
masak mornin 05:01
07:05 cls_bsd__ is now known as cls_bsd
nothingmuch hello 07:17
crysflame . 07:21
scw Hello, any one knows the meaning and rule of :key{ a => 1 } ? 07:23
nothingmuch yes
that's a pair
scw e.g. what should it produce? can there be more than one pairs in {}? 07:24
nothingmuch whose key is "key"
and whose value is an anonymous hash
with one pair inside it
(a => 1)
S04 discusses disambiguating blocks from hashes
scw so :key{ a => 1, b => 2 } is also valid?
nothingmuch yes
it's really the same as key => { a => 1, b => 2 }
or :key{ :a(1), :b(2) } 07:25
scw great!
nothingmuch: thank you vry much :)
nothingmuch you're welcome =) 07:26
this pair syntax is everywhere, btw
double angle quotes know how to use it
for example, enum <<:a(4) b c d>> means:
scw I saw many :key('val'), :key<val>
nothingmuch <val> and ('val') are equivelent 07:27
back to the enum:
the quotes make a list:
((a => 4), 'b', 'c', 'd')
the enum gets that list
and sees that a is a pair, so it enumerates 'a' with the value in the pair 07:28
and then ++'s the value
scw make things quite clear
nothingmuch =)
for each element that isn't a pair 07:29
so b gets 5, etc
scw btw, is :key(1, 2) also valid? I guess it should be 'key' => [1, 2] ? 07:30
nothingmuch i think yes, but i'm not 100% sure
luqui should probably know 07:31
if $a = (1, 2) puts an array reference into $a then yes
looks like yes 07:32
pugs> my $a = (1, 2);
pugs> ref $a
'List'
it's pugs> my $b = [1, 2];
pugs> ref $b
'List'
scw Hmm.. so perl 6 is intuitional and implementation contains many reuses :p 07:36
luqui I should know what 07:38
luqui is reading
nothingmuch about how lists scalarify 07:39
nothingmuch hasn
't yet read those parts thoroughly
luqui Yeah, pretty sure :key(1,2) means :key([1,2])
scw luqui: And are :key('val') :key<val> :key{ k => 'v' } all the forms using 'key' as key? 07:43
luqui: besides from key => 'val' and alternatives like :key .('val)
luqui hmm... 07:44
I wonder if :key[stuff] is valid
for certain values of stuff
ahh, :key[1,2,3,4] is valid 07:45
autrijus cool.
scw equivalent to :key(1, 2, 3, 4) ?
luqui Yeah...
autrijus luqui: you've seen my IMCC.hs plan? thanks it's sane? :)
luqui which begs whether one of those should not be allowed
autrijus s/thank/think/
well, :key[1] is explicitly a list 07:46
luqui haven't seen it
autrijus while :key(1) is an element
probably makes sense
luqui yeah
scw agree
autrijus: I can't find out how (1, 2) is parsed in pugs!
autrijus luqui: oh ok. the plan is to ditch the Eval.hs (pugs evaluator)
scw: it's just a binary op "," I think, no? 07:47
luqui good, I never understood that part of it anyway :-p
autrijus luqui: and merge Compile.Haskell (haskell compiler) with Compile.Parrot (IMCC compiler)
the new architacture will be
Perl 6 Source comes in
we parse it to Pugs AST (ASTh.s)
(AST.hs)
then we compile it to IMCC AST
(IMCC/AST.hs) 07:48
then we translate that into Template Haskell
then we evaluate the haskell code.
using Haskell itself.
luqui IMCC is referring to parrot's IMCC?
crysflame hmmm
autrijus the haskell code will just be a IMCC emulator
crysflame neat
autrijus re, I mean IMC (aka PIR)
luqui oh..
scw Oh, listSyn " , "
luqui so you're going to write a pir evaluator in haskell? 07:49
nothingmuch why don't we just depend on parrot while we're at it, and spew things to it directly, via embedding or whatever? 07:51
luqui I think that's a better idea than trying to emulate (ever-changing) pir
autrijus right. 07:54
not quite pir evaluator
pir->to->nativehaskell compiler 07:55
which can be compiled and evaulated on the fly
luqui do you realize how many parrot ops there are?
autrijus pir-relatd ones are few.
it's just Perl*.imc
no?
Perl*.imc and Continuation and a few primitives
don't need to be complete 07:56
luqui opcodes, not pmcs
autrijus should I care? :)
luqui (maybe I'm not understanding)
autrijus I'm not implementing a pasm cpu
I'm merely implementing the small subset of pir/imc that is neccessary to run pugs
luqui oh.. right
autrijus nothingmuch: because: 07:57
1) embedding parrot is very hard at the moment.
(mainly because embed.h was not installed and theere's no ExtUtils::Embed like things)
2) a reference IMC implementatino will free us from the dreaded deadlock situation with PIR bugs and shortcomings 07:58
3) We can then understand the operational semantics of IMC without having to trace .c code.
4) The tests can be shared between IMC and Pugs.
luqui are you going to write a register allocator too?
or just stick everything in separate registers 07:59
nothingmuch i think we don't care about that
unless we want performance
parrot is supposed to be smart about mapping to real CPUs, right?
luqui yeah, i realized that at some point between my two comments :-)
autrijus right. 08:00
I don't care :)
nothingmuch thinks it could be an interesting way to learn IMCC
autrijus $I* is just a (Map Int Int)
nothingmuch how complex is it, btw?
autrijus and $S* is just (Map Int Strnig)
nothingmuch: extremely simple :)
nothingmuch so basically pugs will be:
luqui wonders, if it's so simple, what parrot has been doing for three years 08:01
nothingmuch parser -> our p6 AST -> lazy compiler for IMCC -> evaluator for IMCC
luqui: i think parrot's focus has been robustness and correcness and portability and efficiency, no?
also, the design is not simple
you have to theorize on how things will look
and then abstract them 08:02
and then break them down, discover their common roots
and design a language to take care of that
plus there's some implementation details that are easier in haskell
GC, and threads, and whatnot
luqui like the fact that GC is already taken care of
nothingmuch ooh
do we have a GC test?
nothingmuch reaaaally wants it to be a bug for things /not/ to be destroyed as soon as possible 08:03
it's one of the most useful things in p5,
luqui nothingmuch: no luck
parrot has to do it then, too
and parrot probably won't
nothingmuch ?
i thought parrot has means for smart and refcount GC 08:04
and you get a sort of hybrid system
luqui no refcount
nothingmuch oi vey!
luqui (unless there have been new developments recently)
nothingmuch so much for compatibility
how are they going to port Ponie?
luqui by hijacking perl5 somehow
nothingmuch that's just plain stupid
GC can be overkill 08:05
crysflame ponie replaces SVs with a PMC-based SV
nothingmuch refcounting is fast
and good
crysflame port it to what?
nothingmuch and enough 80% of the time
crysflame goes back to listening.
luqui according to the studies that Dan read, apparently, refcounting is not fast
(and /me really wants to be able to use circular references liberally) 08:06
why is everyone so attached to timely destruction?
nothingmuch timely destruction helps me manage resources
it's a way to do cleanup
but encapsulate it
instead of saying $obj->im_done
you just throw it away
and $obj takes care of itself 08:07
luqui how have you used that?
nothingmuch i have a big system that runs tests on remote machines
i have resource allocation for things like remote shells, semaphores representing remote resources
daemons, work directories, etc
it's not a big change to add $obj->im_done
to the superclass that implements the job stealing resources from people
but it's yet another thing i don't want to worry about 08:08
luqui i understand completely
luqui just had to write cleanup() callbacks for Logic.pm
and it's a pain, even though it's not
nothingmuch and /maybe/, /sometime/, perhaps they will be destroyed
it's more of the horror stories from the other side that scare me 08:09
lumi said something along these lines about java:
you have objects, and err, then you're done with them
he said it was very annoying
luqui well, perl 6 has lexical hooks, which takes a good 80% off of the need for timely destruction 08:10
but there's still that 20%
nothingmuch wonders why a hybrid system can't be formulated 08:11
luqui because Dan has been opposed to it
and leo has been silent about it
but maybe Chip will listen :-)
leo has been trying to work in a generational gc for awhile
luqui hasn't really followed that
nothingmuch a hybrid system pleases both sides 08:12
you get "proper" garbage collection
which takes care of loops, in some smart way
and some kind of reference count like system to invoke small increments of that thing when things are thrown away
luqui you know that you can never get timely destruction on circular structures 08:13
so what if people start expecting timely destruction, and they have a circular structure, and then it doesn't work
as opposed to saying up front "there is no timely destrution" 08:14
if we say "there is timely destrution most of the time"
then bugs will be subtler
nothingmuch what if a generational GC was triggerred by a refcount system? 08:15
or at least a container destruction system/ 08:16
i mean, say a block exits
it's lexical scope is thrown away
luqui scw, to get back to you about an hour ago: there is also :key<<value1 $value2 value3>>
nothingmuch the generational GC can then treat the young generation as the stuff that was referred by the pad
luqui because you still need to do a full sweep to guarantee anything 08:17
otherwise you're just guessing, and getting it right most of the time
but not all the time
nothingmuch what about introducing mutual weakness at the reference level?
trying to detect what is circular
and when one of these links is traversed, it and it's siblings are weakened 08:18
luqui I wonder if there's a good way to detect when you have circular structures
scw luqui: How about unicode version of << >> ? they should be equivelant?
nothingmuch detecting circularity in this manner shouldn't be too complicated
luqui I was actually referring to the unicode versions, but I believe they're equivalent everywhere 08:19
the question is, can you do it efficiently enough
nothingmuch is really not familiar with graph theories
should go learn compsci some time
luqui the problem is that most of these "detections" are O(n) 08:20
and n is all the objects in the program
which brings you right back to what we have, mark-sweep
nothingmuch www.cs.cornell.edu/boom/2004sp/Proj...dex16.html
brentdax My God, you Pugs people have been busy.
nothingmuch that's interesting, though it doesn't solve our problem
luqui I've hard of that
heard
Kim wrote an article about it 08:21
(Kim Burchett, of kimbly.com)
(one of Dan's favorite blogs, it seems)
heard it was very awkward
nothingmuch how do ddd and other similar programs work? 08:22
luqui I think they annotate the memory allocation structures 08:23
and then test whether any are remaining at program exit
like valgrind
A friend points us here: dev.perl.org/perl6/exegesis/E07.html 08:24
er not there
here: www.cs.colorado.edu/~lizb/private/p...diwan2.pdf
lightstep some early smalltalk systems had a hybrid system: you had refcounts, when they overflowed the object was delegated to the gc, and once in a while they did a full sweep 08:25
nothingmuch citeseer.ist.psu.edu/context/13435/2716
lightstep but this scheme wasn't good at all
luqui what were the weaknesses? 08:26
nothingmuch can we use typing to make GC sweeps less slurpy?
or does parrot not support typing?
autrijus sorry, bbiab :)
nothingmuch ciao autrijus 08:27
luqui well we certainly can't rely on it
lightstep luqui, it was slower than generational gc
luqui and most of the languages we support wouldn't give it any information
nothingmuch but in theory a language compiler that is smart enough could help the GC system be quick 08:28
lightstep nothingmuch, i don't understand. are you worried about memory leaks or about activation of finalizers?
luqui it started as the latter :-)
nothingmuch both
=) 08:29
i prefer activation of finalizers
but most people prefer being trigger happy with circular structures
my style for increasing robustness is to put all annoying details together
in the begining
that is, all refcount weakening is centralized
for the same object
unless I really can't pull that off
this lets me rely on refcounting to stay pretty
and for me it's worth the effort
luqui have you seen what POE has to do to get around the circular structure bs? 08:32
it has to implement its own refcounter
nothingmuch guesses his programming style will have to change
ook
i've done that once
i wanted to have an LRU cache that purges at O(1)
ofcourse, it didn't, since it was using perl structures 08:33
but it was just an excercise
and when it's accessed it's spliced up to the top
it had a doubly linked list, where references go in
the hash that caches contains a reference to the link element 08:34
when the cache overflows elements are popped off the list
nothingmuch loves writing things and feeling really clever, when he tries to understand a feature
since then, when I learned about refcounting, and made myself face it the hard way, i like it
because what it gives me, it gives at a cost i'm willing to pay 08:35
luqui basically, which cost is the general programming public willing to pay?
lightstep the common wisdom is that gc is at least as fast as refcounts (less page faults)
luqui learning about refcounting, or calling .close on things 08:36
yeah, cache locality is a big issue, so I'm learning while working on parrot
nothingmuch Cocoa has a
weird feature
they call them allocation zones
luqui go on 08:37
nothingmuch which are a techinque to help optimize the allocation, so that related objects are paged from the swap together
lightstep oh
there's talk about regions in dome papers, but i don't know what they are
luqui is cocoa refcounted? 08:38
lightstep maybe it's related
nothingmuch so the NSObject base class has [class alloc] and [class allocWithZone:Zone]
kakos Obj-C is
nothingmuch cocoa is refcounted manually
luqui ok
nothingmuch kakos - objective c isn't, it's just C
kakos nothingmuch: Really? My bad then.
nothingmuch but you have a shortcut
cocoa is, by supplying the allocation pool objects 08:39
anyway, what you have is a sort of singleton
objc/cocoa are very tightly coupled though, so it seems like it =)
except it's really a stack
you push an allocation pool onto it
luqui are you really typing that fast?
nothingmuch and then you tell objects to autorelease
and they say to the release pool, destroy me whenever
and when the pool is destroyed it refcounts it down
yes, i am ;-)
luqui you type faster than I read
nothingmuch i type too slow, IMHO, or rather, I think that I think too fast 08:40
i need to type slower, and think much slower, and then maybe i could make some sense
languages have never been my friend
luqui well, I've wasted far too much time 08:41
here
"wasted"
nothingmuch ditto =(
luqui I need to get back to my homework which is due in six hours
luqui & 08:42
nothingmuch good luck, luqui 08:43
www.google.com/url?sa=U&start=4...amp;e=7152 08:44
i think i'm reading that right now, not too sure
it looks like generational meets alloc zones, or whatever
nothingmuch wonders if a zone system with an index of all refererences to and from foreign objects could be used to try and search for circular objects in a more graphical manner 08:55
tightly related objects link within the zone
so to not cause too much overhead
with multiple zoning levels, and circularity between zones dealt with traditionally, you could move the complexity to a smaller N, i think 08:58
the cost being at reference time 08:59
within zones, an optimistic GC could be safer, because it knows what objects are foreign
zones should be per object at the lowest level, and if two zones share a great number of links, then they are bound together in a new zone 09:01
and insignificant zones can subsequently be optimized away
which is really where there is a gamble
since then you can't repartition a zone easily 09:02
lightstep but when creating the program you have some idea of the memory allocation patterns. throwing it away and reconstructing in runtime is bad 09:03
nothingmuch you mean, the object definitions, etc?
lightstep the zones 09:04
the programmer, and even the compiler, have data that can significantly improve the zone tree
nothingmuch right 09:05
lightstep and using this impormation might obviate the repartitioning at runtime
the need for
nothingmuch exactly
with that data
and also statistics
either gathered, saved, and learned
or during runtime, for prediction
could help the GC throw away zones which can be safely forgotten about 09:06
jabbot pugs - 1275 - Testing the wrong thing. 11:57
pugs - 1274 - Accepting :key('val1', 'val2') and :key[
pugs - 1273 - Accepting :key<val1 val2>.
autrijus wow.i 11:58
nothingmuch ?
autrijus :key[] :key<> and :key()
so nice.
obra hey autrijus
autrijus hey obra.
MRT++ # fast fast wireless 11:59
obra nice
autrijus: how'd you solve their issue?
autrijus scw++ # mad and motivated
obra: I had.
clkao svk branched for 1.0 release engineering
luqui hey autrijus, have you begun BEGIN?
autrijus obra: their visual basic component was not freeing its handles.
luqui: no, feel free to work on it.
obra is that the issue?
nothingmuch if it weren't for crazy people pugs would never have been... =)
obra the whole issue?
autrijus luqui: it's part of the 3-stage compilation plan.
obra: well, there's various complications.
obra: I switched DBD::ODBC to DBD::ADO; that fixed threaded contention 12:00
then I increased timeout, which fixed starvation
luqui hmm... I'm thinking the best way to do it is to implement macros
nothingmuch autrijus: give me an idea for something to do for pugs
i'm a bit sick of fighting with vim
autrijus but the core bug is that their VB component was not doing things correctly.
luqui and BEGIN is a very simple macro
autrijus luqui: explain. BEGIN is compilation time, not runtime
luqui fighting with vim?
vim is wonderful
autrijus obra: so I'm like, charging them fully, because that's outside my contractual obligation. 12:01
nothingmuch it is
Text::VimColors is not, though
luqui autrijus: macros are also compile time
nothingmuch for some reason, even though it display =pod .... =cut as comments normally
autrijus luqui: oh. macros with side effects, eh.
obra autrijus: nice
autrijus luqui: sure... I punted macros only because they are not specced at all.
nothingmuch the html from Text::VimColors has the pod parsed as perl
luqui they are a little bit
anyway, should I just use what's in place in Eval.hs? 12:02
autrijus luqui: not to my eyes, so feel free to work on it :D
nothingmuch the mysterious 'is parsed' trait is sooooo scary
autrijus luqui: sure. none of Eval monad will be invalidated
luqui okay cool
autrijus it will just magically become a compiler.
that's the point of this whole monadic revolution :)
nothingmuch we're thinking of forking the test reporting stuff into another project
what should we call it?
luqui nothingmuch: but is parsed is not a macro-only thing 12:03
autrijus nothingmuch: look at test failures and pick some easier ones?
nothingmuch it should do the doc backlinking, the pretty graph things, and whatever
luqui it's quite orthogonal
autrijus nothingmuch: Test::Visualize ?
nothingmuch smoker data consolidation
it's more a collection of scripts
why not
scw autrijus: I just want the function call write "something", :bold 12:04
:p
autrijus scw: and it works now? 12:05
scw autrijus: definitely :) 12:07
nothingmuch wonders where his funky GC scheme is utterly broken
luqui: did you read a bit?
scw autrijus: t/subroutines/sub_named_params.t:88
luqui no, I finished my math test :-)
autrijus scw: wonderful!
luqui nothingmuch: it might be good to outline it in a little doc 12:08
so people can think about it concretely
nothingmuch are you sure that isn't pointless?
it was more of an excercise in triz like thinking 12:09
nothingmuch tries anyway
luqui well, mostly because I have no idea what your scheme is... 12:10
nothingmuch ok
luqui and I'm not paying enough attention right now to listen
nothingmuch so i'll write =)
autrijus needs to work on PA02 too. 12:11
but, taking MRT back home &
gaal nothingmuch: how about Smoke and Mirrors? (S&M for short :) 12:15
nothingmuch gaal: that's not very search.cpan.org savvy
gaal the yaml harness is almost entirely not dependent on pugs
it needs some smarts to detect other source control system
s
nothingmuch what I think we'll do: 12:16
create a test reporter
that creates YAML
and inserts revision/arch info
the YAML harness is used by it, but unrelated
gaal plus there's one place where i get the version of pugs itself, but if there's a contract that the project declares its version somewhere clear, i could use that
nothingmuch then we have data consolidation
i think YAML harness should take that data out, and deal only with actual test results for now 12:17
gaal "test reporter" can just be another output mode for Test.pm, no?
jabbot pugs - 1276 -
nothingmuch the catalogging thing should be decoupled too
into a doc/test link system
where you can declare shortcuts for doc names
so we can retain S02
the grapher should be a bit smarter, and take consolidated results as a data structure 12:18
and display them as such
the consolidator takes YAML, and makes a structure for the grapher
and lastly Test::Builder should have the $?CALLER::CALLER::POSITION hack for graph->file linking 12:19
gaal that's a useful extentsion to TAP, in any case 12:21
nothingmuch yah
we should delegate that responsibility to andy lester
gaal petition for it, you mean :) and offer our current Test.pm with our appeal 12:22
leo_ fireall test 12:26
luqui is leo leo?
leo_ leo is leo :)
nothingmuch leo_: pond
luqui hi leo 12:27
nothingmuch s/d$/g
leo_ my first visit to IRC ever - just looking what my firewall says 12:28
obra heya leo
clkao leo! 12:30
q[acme] waves
leo_ yes?
Juerd Hi leo_ 12:33
nothingmuch leo_! leo_! leo_! 12:34
leo_ hi all
nothingmuch mimicks someone elses voice: "is leo_ here?"
leo_ yes - first time on IRC
q[acme] and what does the firewall say?
leo_ not sure yet but modeprobing ip_mask_irc seems have done it 12:35
autrijus hi leo!
I'm in middle of MRT station
it only stops for 10 secs
leo_ hi autrijus 12:36
autrijus so... brb :))
Juerd Hi autrijus, Bye autrijus.
autrijus = be right back
hcchien fg 12:37
ouch.....
Juerd heh
leo_: How have you managed to live without irc? :)
autrijus :) at jian tan station now
err, shih lin
obra hey autrijus 12:38
Juerd I'm 21, and spent the last 10 years on IRC... :)
I can't imagine life without :P
obra we got clkao to branch svk
leo_ well - the metered ISDN line. it already dropped the line once
hcchien autrijus: back to Danshui?
leo_ I'm 47, and spent the last 12 minutes on IRC ;) 12:39
Juerd Heh, I imagined you younger (no offense) 12:40
leo_ I feel younger 12:41
are there any Parrot related questions? 12:42
Juerd Does IRC make you feel younger? :)
I have no questions at this moment
leo_ IRC looks still very strange to me - but I get used to it 12:43
Juerd If you need any help, just ask for it
leo_ thanks, not yet, I've to explore it a bit first 12:44
pjcj Do you know where #parrot is? 12:48
q[acme] pjcj: you scared him off!
Juerd Not there :)
nothingmuch luqui: nothingmuch.woobling.org/gc_zones.txt
i have only thought about it about 10 minutes though 12:49
so it's likely to have big holes
but it appears to me that with a certain overhead when you refer to other objects, you can gain a lot later when looking for objects to destroy
leo_: this might interest you too... ./configure --disable-layout --disable-tests --disable-samples
oops, nothingmuch.woobling.org/gc_zones.txt
though please take it with a grain of salt, as I have no clue what I'm doing. 12:50
integral ah, zones can't form loops so it's safe to ref count them? 12:52
oh, not at all. 12:53
nothingmuch integral: ah? 12:55
zones are a tree of partitions and sub partitions
a zone can have only one parent
and knows about zones it's children relate to
a zone whose children relate to no one
not even the roots of the data tree (registers, or the stack) can be simply deleted as is
integral oh, so the zones nest, rather than partition the whole of memory 12:56
nothingmuch right
if they nest down to the lowest level, then you get nothing
the real issue is to know what zones don't help you
ideally many levels of leafs
so that your low level zones are several hundred objects, maybe a little less, maybe a little more 12:57
integral: am I being too naive?
integral I don't really know enough to tell. How are the low-level zones being built? 12:58
Are they created from compile-time information?
nothingmuch well, in theory each object has it's own zone when it's constructed 12:59
and when it reaches out to other objects it should have it's zone optimized away very quickly
in practice you could pre optimize, maybe
compile time info helps you partition the zones more wisely, methinks
by using type constraints you can know where zones won't touch, because their objects can't reference each other
the hosted language could provide this data to the GC
Juerd Holy crap, this channel has 86 nicks in it
nothingmuch Juerd? 13:00
Juerd nothingmuch: I don't know any other channel that has grown this fast
nothingmuch err, not provide
but help the GC
ah
it grows by about 5 people a week, on avg, afaict
Juerd No, much faster. 13:01
nothingmuch it seems like a lot, but then again, i have never lurked on IRC
Juerd It has existed less than two months.
nothingmuch i mean over the last 3 weeks or so, since i joined
integral I don't know any project that has gathered commiters so fast either :-)
tombom the only reason i came was because of ghc 6.4 reporting an impossible error in main.hs
nothingmuch that is what's really impressive =)
Juerd integral: yea
nothingmuch (to me)
so, does anybody have on opinion about that GC scheme? 13:02
leo__ line dropped again :( 13:05
Juerd leo's collecting underscores 13:06
nothingmuch i was just about to make a joke about leo___ 13:07
leo__ I don't yet understand gc_zones.txt
yeah, what are these underscores?
nothingmuch when the nick is taken
since the old you did not log out yet
Juerd leo__: When you reconnect and the old client is still "connected" to the server, the old nick is taken
leo__ by me
nothingmuch then you get a temp
Juerd leo__: Your client then has to make up another nickname, usually formed by ~= "_"
nothingmuch anyway, the zones basically are: 13:08
Juerd leo__: Use /nick leo to return to the original one
nothingmuch a way to group objects into groups
for the sake of simplifying grabage collection
so that it's feasable to do something equivelent to a full sweep 13:09
in a complexity closer to reference counting
except that it's a bit more complicated than reference counting, and possibly (probably) completely wrong
leo__ one group per object kind?
nothingmuch that's how it starts
skypp #haskell
Juerd skypp: Type /join #haskell to join that channel
nothingmuch but very quickly the leaf level zones should be forgotten
Juerd skypp: Without the /join, it's a normal message.
nothingmuch and the objects close to each other just packed directly under one medium sized zone 13:10
lumi "delete whole zones at a time" -- In what order are the objects destroyed?
nothingmuch lumi: =P
lumi Frankly I'll just be glad for finalizing at all... :P
nothingmuch where do you start to destroy a cyclic structure? 13:11
at the node that was first refcounted to zero?
brb, i've got to rescue someone
lumi True, but that's not an answer
nothingmuch i don't know 13:13
lumi I guess they'd need to order themselves if it's important
nothingmuch i think that's for the language design to decide
or provide hooks for
in theory you have a set of object
s
unordered 13:14
in a zone
you could destroy based on reverse chronological order of creation, if that's your thing
i dunno
leo__: is it any clearer?
lumi: err, i've been doing a lot of 'echo powerpc-apple-darwin7.8.0; exit' in config.guess files 13:15
leo__ not much, no - need probably a picture of object/zone relations
nothingmuch think of a graph of objects
and zones are non intersecting circles
which could contain other smaller circles
theorbtwo Mornin, all. 13:16
nothingmuch morning theorbtwo
leo__ I think a generational GC system has to tightly integrate with the interpreter
e.g. it has to know, when a scope is created or destroyed 13:17
nothingmuch but generational GCs can't ensure immediate finalization, right?
leo__ we have to do a lazy GC sweep at scope exit 13:18
there are two cases: eagerly object is found - fine
or not then the next older generation has to be considered 13:19
theorbtwo Is it just me, or is there a sharp increase in test fails?
nothingmuch theorbtwo: i thought there was a decrease 13:20
leo__: is a generational lookup equivelent (in terms of side effect) to a full seeep? 13:24
swwp
sweep
damn
leo__ if e.g. the ParrotIO object is found in the youngest generation, only this generation is looked at 13:26
nothingmuch is a bit slow
could this means that a now unreferred object exists in an older generation?
like, if i explicitly 'undef' a global?
or is that special case treated? 13:27
leo__ it depends. If someone was stuffing the IO object into an array, the generation of this arya has to be swept
that cab of course get expensive
nothingmuch what my idea sort of tries to handle is to allow a reference counting like approach for finding data 13:29
but with mark & sweep safety/correctness
by doing some cross indexing when objects refer to each other
leo__ ref counint is nasty, especially with shared objects 13:30
luqui nothingmuch: re gc_zones, your system seems to incur too much overhead for what we're looking at (people complain that java programs are memory hungry, we'd rather not say the same) 13:32
but it does bring up an interesting idea
nothingmuch luqui: ofcourse, it's a question of balance 13:33
i think the overhead is big
luqui could we refcount on topological closures of objects
if we could, then we could handle circular references
nothingmuch but if you optimize your zone tree like btrees are 13:34
with wide nodes (wide zones), in theory it could work
explain please =)
luqui do you know what a graph closure is?
nothingmuch no clue
nothingmuch only knows what 'edge' and 'dot' means
luqui it's the set of nodes, starting with one node, that that node "eventually" points to
nothingmuch oh, node is the term for dot =) 13:35
luqui hehe
theorbtwo Yup.
nothingmuch how do you generate this closure?
luqui using the closure algorithm
Limbic_Region leo does IRC now?
luqui basically a simple breadth-first search
Limbic_Region welcome ;-)
luqui for instance, the LR parser makes closures of productions in order to figure out what states it needs 13:36
what if the GC objects were object closures?
I'll have to think about this one a bit
it might be really cool
(and will probably be too expensive to do in practicality)
nothingmuch luqui: i think it's sort of the same 13:37
luqui yeah
gee, I wonder why I thought of it right after reading your description :-p 13:38
Juerd Speaking of closures, can join '', 1..5 => map { foo() } be written as { foo() } x 5?
nothingmuch =P
luqui Juerd: it's an interesting proposition
but then we'd probably have to do (the much more commonly wanted) { foo() } xx 5 13:39
Juerd Both.
luqui which would prohibit us from duplicating a closure that way
but it might be a worthwhile trade-off
Juerd Why would you want 5 times the same closure?
luqui why would you want 5 times the same string?
nothingmuch luqui: what i think i tried to do, if i understand the closure idea, is to attempt to optimize some of them away
Juerd NOT SAME! :)
That's the important thing.
It should call foo() 5 times, as the map does 13:40
luqui I'm just saying that (list)x5 was in Perl 5 for a reason
Juerd Yes, I realise that
nothingmuch and use the statistics collected to create an approximate representation of them, while the references are built up
Juerd But I can't think of a reason for needing duplicate closures yet :)
luqui well, closures are objects too
nothingmuch Juerd: to trick something that thinks it's taking distinct closures
Juerd And if you need them, the "sub" keyword can probably disambiguate 13:41
luqui I think if we do it purely on the syntactic level I think it should be okay
my $foo = { bar() }; $foo xx 5; # five of the same closure
Juerd Or even double { }
luqui { bar() } xx 5; # call bar() five times
Juerd { { foo() } } xx 5; # 5 closures
sub { foo() } xx 5; # 5 closures
nothingmuch how do you tell apart a closure with a closure
luqui Juerd: those would all be distinct (in your earlier example) 13:42
Juerd { foo() } xx 5; # whatever the 5 foos returned
nothingmuch { { } } from { }
at the user level?
Juerd nothingmuch: What is 'user level'?
nothingmuch using perl6
how could i know if closure x is { { } }
and closure y is { }
where is the difference? only at the syntax level?
Juerd You don't have to
nothingmuch if so, then it isn't reusable
luqui I think he's saying that if you execute the closure, you get another closure back
Juerd { { } } benefits from executing the outer closure, returning the inner.
gaal how do i declare a class constant? our $F_STORABLE is constant = 1; ?
nothingmuch if there is a difference, then i think it leads to inconsistency 13:43
ah
ok
i get it now
luqui but the closures you get back would all be distinct
so that's a problem
Juerd (Note that I haven't read on hash/sub choices yet)
luqui: Good point.
nothingmuch so basically what you want is for 'x' to interpolate /after/ it's repeated?
luqui So like I said, it's probably okay if we do it on the syntax level rather then the semantic level
nothingmuch i disagree with that 13:44
autrijus leo_: greetings!
luqui nothingmuch: por que?
autrijus I'm now safely back to home.
nothingmuch it's not p6ish in the consistency approach
it's a special case
"why was this called 5 times?"
luqui Perl 6 has syntactic special cases
Juerd nothingmuch: No, I want a better syntax for the extremely often used map 1..x where 1..x is just to make it loop x times
luqui why does #line give me metainformation?
nothingmuch "well, if you read chapter 54.3.2.5 you would have seen that"
luqui: but we want to minimize them, right?
Juerd nothingmuch: See ~~ - it has syntactic matches.
luqui yeah, that's true 13:45
Juerd nothingmuch: These you can't pass as values, for example
But they are useful.
nothingmuch isn't that better written as:
~({ foo () } xx 5)?
instead of { foo() } x 5? or { { foo () } } x 5? 13:46
luqui but that's the feature we're discussing
whether that closure would execute
or whether you'd just be trying to stringify the closure
nothingmuch i think the more general the case, the more desirable it is
luqui ?
nothingmuch you could express { foo() } x 5 in Juerd's way
even though it's typical
with perl6 semantics that have no special case 13:47
Juerd nothingmuch: I think the more practical use something has, the more desirable it is
nothingmuch at around 3 chars more
Juerd: but then you block out a certain feature
or require adding a special case
Juerd Perl stands for dwimmery more than for consistency.
luqui nothingmuch: your ~({foo()} xx 5) is not standard perl 6
and would be a special case
nothingmuch why is that a special case?
luqui because currently that would not execute the closure
Juerd nothingmuch: No, the other feature would require syntactic disambiguation, for example by adding a "sub" keyword.
nothingmuch Juerd: but reading that for the first time 13:48
without having read about the special case
does not make any sense
it's like pascal
luqui Juerd: I think adding a "sub" keyword is the wrong way to disambiguate, but I agree with you in principle
I think it naturally falls out if we allow {foo()} xx 5 13:49
nothingmuch inconsistencies with pascal that really hinder readability:
in declaration you have 1 .. 5, in loops you have 1 TO 5
begin ... end; are symmetric, except where they aren't
some block openings are two lines, others are one line, or else they don't match up well
Juerd This isn't pascal
nothingmuch var is used too often
to declare the property of a name
and it's behavior 13:50
and a section where names are declared
luqui okay, I think you should go to #pascalrant
:-)
nothingmuch these are things that perl sometimes exhibits
though less obviously
which is even more dangerous
luqui then give us perl examples
nothingmuch throwing a 'sub' to disambiguate is a bit yucky
Juerd It's a problem if the result is hard to spot
nothingmuch luqui: using parens and stuff to listify in p5 13:51
() = ....
() x 5 in list context
Juerd nothingmuch: I agree that 'sub' is probably not the right way
nothingmuch etc
Juerd It's the best thing I can think of rigth now.
right
nothingmuch i really want this stuff to not exist in p6
Juerd We're not discussing P5 either
nothingmuch so the language will be less 'well, you should have used this magic construct'
Juerd Without magic constructs, we end up with python, nothingmuch 13:52
That's not Perl.
nothingmuch the power of hebrew, for example, is that you can construct logical parts of the language
with special cases that are obvious
and find out how something will look like
and understand something we haven't met yet
luqui and "obvious" is the subjective word here
nothingmuch by means of a thought process
luqui most everything in perl 6 is obvious to larry
nothingmuch and not only by looking at a reference doc
Juerd Maybe we can introduce X and XX
nothingmuch again, my point is still about arbitrary use of a keyword with a /lot/ of meaning and cargo, to disambiguate a special case 13:53
without some logical binding
Juerd Where @foo xx 5 is 5 times \@foo, and @foo XX 5 is foo's elements, 5 times
and { foo() } xx 5 is 5 times that closure, and @foo XX 5 is 5 different return values of foo() 13:54
nothingmuch perhaps for consistency it could be made *x
i don't see myself, as a newbie saying "oh, since the sub keyword is used, then it doesn't get executed till after it's repeated"
Juerd: i like that much better
luqui I p6l'd the discussion
nothingmuch * has a strong link to flattenning vs. referencing
Juerd nothingmuch: I understand the problem you have with the syntax - but please note that I'm more suggesting a feature with some syntax to demonstrate, than I'm actually proposing a syntax change
nothingmuch Juerd: ok, sorry if i was a bit too nasty 13:55
luqui Juerd: then you might want to reply and elaborate to my p6l message
if you have any other ideas, that is
rgs are there q// strings in perl 6 ? besides qq// 14:00
luqui yeah
rgs not implemented yet, then
luqui yeah they are 14:01
(I did that... unless they're broken)
leo_ I'm at #parrot now - bye
luqui I may have only done qq
nothingmuch leo_: #parrot is not on this network, btw
so are you on freenode at all?
luqui yeah, looks like I only did qq
Juerd leo_: With a good IRC client (I suggest irssi), you can be connected to multiple servers and be in multiple channels at the same time.
leo_ on freenode and pobox - yes
thanks I'm using irssi 14:02
Juerd Ah, you use irssi
Then you can use /connect instead of /server
And it'll *add* a connection, rather than replace it.
leo_ yep
Juerd Every window gets a number, meta+digit switches between the first 10 windows
luqui (something tells me he knows how to use irssi) 14:03
Juerd (meta being alt or escape on a pc terminal)
luqui: That'd be weird though, given that it's his first time on irc :)
scw qw// is in pugs, too
lumi Possibly spider sense, or tfm? 14:04
luqui with closure interpolation and all
scw if you have more than 10 windows, (quite usual, isn't it :p), meta-q meta-w ... is used 14:05
luqui and more than 100 is ^Zrm -rf / 14:06
Juerd luqui: replied
The string x is more important in this respect, I think
Because join+map communicates the wrong thing
luqui agreed 14:08
well, now it's in larry's brain
Juerd heh
luqui (he's been holding forth, so maybe he'll have some insight about it)
any('a'..'z')x5, eh? 14:10
well, I think map { $_.pick } any('a'..'z')x5 would work
nothingmuch damnit, someone is spamming as me
luqui and now it's communicating the right thing
s/x/xx/ 14:11
gaal i just ci'ed a small patch to util/perl6.vim to support $:x and %.y, luqui and all please see if this makes sense to you. 14:40
beh, $work. bbiab 14:42
Juerd I think I've just gone mad. 14:44
(See p6l)
afk # typing hurts 14:45
autrijus gaal++ 14:46
autrijus is still brainracing with leo on #parrot.
Alias_ OH DEAR GOD NO!!!
More operator modifiers! :(
My brain should not have to hurt in order to add numbers to a list 14:47
jabbot pugs - 1277 - Treat $:x and $:.y as variables
integral hmm, I see the FiniteMap in Config.hs isn't exposed to perl6 code at the moment, would it be inappropiate to expose it as %*CONFIG?
autrijus it needs to be exposed separately as $*FOO
I mean $?FOO 14:48
since it's determined at compile time
but please do
integral ok, I'll have a go at hacking that. Corion was asking for pugs version info to be accessible from perl6 you see
autrijus danke
Alias_ sobs 14:49
rgs ultimately, this would be useful to know whether some code runs under pugs or another interpreter
Alias_ On of the things I really like about perl is that I'm so much less likely to get RSI than when writing Java... 14:50
Juerd Alias_: The upside is that you can just combine operators instead of remember many
Alias_ Perl 6 is going to change all that
PerlJam Alias_: how so?
Juerd Yes, there are many bitwise operators
But it's really just a very simple table of things you can combine
Alias_: Haven't you seen how much typing ~- and [-] save you? 14:51
Alias_ PerlJam: Because the contortions I'm going to need to go into in order to do anything...
Juerd Adding features doesn't give you RSI - using them when another solution is better does
Alias_ PerlJam: $#$%#$...%%#$name = 'Alias';
PerlJam Alias_: I don't see how it will be any different in perl6 as compared with perl5
Juerd Alias_: That's $ and a comment ;)
In fact, it stays meaningful syntax longer in Perl 5 than in Perl 6 14:52
rgs I'm trying to add q/.../ to pugs
Alias_ PerlJam: !@#%@#!%#%!#lastname = $#$%#$...%#$name ~~~~~~~ 1; # <--- magical do what I mean operator
Juerd Alias_: No, that's called >>dwim<< :)
stevan Alias_: why is the double sigil so distasteful to you?
Alias_ stevan: Because something shouldn't have to matter...
PerlJam Alias_: for instance, perl6 cleans up perl's "cognitive space" by being more verbose in some areas, but you can easily alias the verbosity if RSI becomes a problem ;-)
Juerd I like secondary sigils, but think they are used too much. 14:53
Alias_ It's forcing people to learn things way too early
PerlJam Alias_: I disagree.
stevan Juerd: but (at least IMO) they seem to be used very consistently
Juerd For example, I'd like to just have one object with lots of methods instead of many $*VARIABLES and $?VARIABLES
stevan: That is true.
PerlJam Larry has jumped through hoops to *not* force people to learn too much too quickly.
Juerd stevan: If $*PID is an alias for %*<PID>, it's fine by me, but I'd like to be able to dump these things all at once. 14:54
afk
stevan Alias_: it seems to me that they follow very basic and consistent rules (but I have not read all there is too read on them, so I may be wrong)
Limbic_Region or use a slice to get just the ones I want (all at once)
Alias_ stevan: But that rule involves learning
There are 5 different twigils yes?
stevan Alias_: whats wrong with learning :)
PerlJam Alias_: give or take.
Alias_ stevan: Because learning is _hard_, and it's a bad thing to increase the learning curve 14:55
Juerd PerlJam: Heh, both given and take are new keywords :)
stevan Alias_: perl6 will involve learning no matter what
PerlJam Juerd: Lots of opportunities to pun in perl6 ;)
Juerd Alias_: Then why are we even creating Perl 6, and not just using Ruby instead?
stevan Alias_: it seems to me that the learning curve just has some steep plateaus in it, thats all
Juerd PerlJam: take my $wife = 'drunk';
stevan Alias_: I think they will get smoother over time 14:56
Alias_ I hope so
stevan Alias_: its still early
metaperl lol @ juerd "why are we even creating Perl 6, and not just using Ruby instead?"
Limbic_Region Alias_ - I am not sure I agree with your original sentiment
Alias_ I'd rather not have for someone to learn about the zillion execution phases in order to just get access to the PID
Limbic_Region given that I learned p5 less than 3 years ago, it is still fresh in my memory
Juerd Ruby is Perlishness without learning curve
And it is rather boring.
PerlJam I haven't decided if people who know perl5 will have an easier or harder time with perl6. I hope it's the former, but there may be too many subtle differences.
stevan Alias_: thats what Ponie is for :) 14:57
Juerd It does have minus for arrays though :)
Limbic_Region I was able to accomplish what I wanted to do (which was beginner stuff) with only a few hours of learning
I very much expect p6 to be the same way for most
metaperl Perl6 is going to lose it's human feel.
stevan PerlJam: I think people who know perl5 AND some other langauges (java, lisp, etc) will find it easier
Juerd Alias_: Have you ever needed to learn the hooks tied to %ENV, or even that %ENV is a hash, in order to use $ENV{PATH}?
You see one thing and copy it
And when you have used many similar things, a pattern begins to form in yourmind 14:58
Limbic_Region on the other hand - once you get past the accomplishing 90% of the tasks to the remaining 10% - you could spend a lifetime mastering it ;-)
Juerd And then you have learned rules you can apply to more things.
Alias_ Juerd: "It's a magical HASH"
Juerd Alias_: But you don't have to know that in order to use it.
integral and from there people are copying perl5 like "&func(...)", and they then wonder why "&un('string')" doesn't work
Juerd Alias_: In the most basic form.
Alias_ Juerd: But now we need to know that it's a compile time variable?
Juerd A bash scripter who learns Perl first tries $PATH
And then asks why that doesn't work. He gets told $ENV{PATH} 14:59
PerlJam why does it suddenly sound like everyone is saying that perl6 will be better for cargo cult programming than perl5? ;)
Juerd At that point, he may not already notice that all environment variables are in that hash. After all, bash makes local and environment variables look exactly the same
Only after having seen another $ENV{HOME} like variable, the pattern is clearly visible
Without reading any manual
metaperl Larry Wall said: "Perl was designed to grow in between shell and C" --- Perl6 is clearly not in line with that philosophy 15:00
Juerd Alias_: No, and for that I agree secondary sigils are a bad thing.
Alias_ But now it's %*ENV?
luqui metaperl: Perl 6 is designed to grow in between Perl 5 and infinity (and beyond)
Juerd * Because it's special. I don't like the ? secondary sigil and think it should all be just *
Alias_ %?ENV?
Juerd $*PID, $*SUB, $*OS
luqui Yeah, now it's %*ENV, which can usually be abreviated %ENV
Juerd &*SUB that is
eh 15:01
&?SUB
luqui Juerd: the $? stuff is always compile time resolved
Alias_ Wait... so if I leave off the twigil, it will just do what I mean?
rgs guys, I've an implementation of q/.../ and '...' that works ; should I commit ?
PerlJam Alias_: yes!
Juerd luqui: Great, but as alias said: why does the programmer have to care about that?
luqui Alias_: in *'s case
rgs and should I remove the old '...' handling code
Alias_ what about in other cases
stevan rgs: does it past the quoting tests?
luqui Juerd: dunno, so he doesn't say &?SUB := sub { ... }
rgs stevan: it solves single_quoted_strings.t
stevan rgs: thats good 15:02
luqui more like, $? stuff represents the environment the program was compiled in
Juerd luqui: That can be an error regardless of the secondary sigil.
PerlJam Alias_: In other cases the twigil means something that you can't get any other way.
luqui and $* stuff represents the environment the program is running in
Alias_ luqui: Who cares? Why force it in people's faces?
PerlJam luqui: and $?PID will do what exactly? ;)
luqui and that's a difference the programmer can care about
PerlJam: $?PID will do "undefined variable"
Alias_ luqui: "can" or "must"
Juerd Can care about and then look up in documentation 15:03
integral why is $* necessary? Can't you just replace $?PID with BEGIN { $*PID } ?
s/\*/?/
Juerd There's no specific reason these things need to be reflected in code
luqui $?PID doesn't mean anything!
Juerd They have nothing to do with the code
They have nothing to do with the program you're writing
luqui what about &?SUB?
PerlJam integral: $?FOO is compile-time, $*FOO is run-time. You don't see a need to differentiate?
Juerd Where the PID comes from is utterly irrelevant, as long as you get the PID.
PerlJam: Is there any FOO that exists in both then?
integral PerlJam: not when the former can be obtained by adding a BEGNI{} block 15:04
luqui okay, forget about environment stuff. we might encapsulate those in objects somehow...
Juerd luqui: I think the entire .* and .? thing needs to be objects
Alias_ If you _really_ care which it is, and that is a rare case, then you get a special method in a special package
luqui Juerd: so tell me about &?SUB
Alias_ Scalar::Util::is_compile_time_variable( $*FOO )
schoch arn't objects sometimes slower? 15:05
Juerd luqui: $PROCESS.PID, %$PROCESS.ENV, &$PROGRAM.SUB
Alias_ or something of that nature
Juerd luqui: $CONFIG.OS
PerlJam integral: Hmm. I think I agree with you.
luqui schoch: quiet. we're language designers. :-)
PerlJam schoch: yeah, and sometimes their *faster* too
integral PerlJam: it seems a bit more perl5ish to just use an existing feature than to invent something specific
luqui I definitely agree with $CONFIG.OS, but I think $? is still necessary to reflect the lexical environment
Limbic_Region schock - p5 objects are almost always slower because of the way they were shoe-horned in
Juerd luqui: That is, if we really need to note where things come from
integral &?SUB could be BEGIN { $*compiler.get_current_subroutine }
Juerd luqui: For all I care, $PID, %ENV, &SUB, $OS are more than good enough. 15:06
Limbic_Region they are taking more of a central role in p6
schoch i'm a java user, one of the things I liked about perl is being able to chose an object model when i need one, or not
Juerd Add a secondary sigil to avoid clashing with your own variables
luqui integral: ahh yes, and it was introduced to make anonymous recursion easy. there that goes.
Juerd $*PID, %*ENV, &*SUB, $*OS
luqui Juerd: and %MY?
PerlJam integral: Well, I'd expect some sort of hook to grab the entire compilation environment in one go though.
Juerd luqui: That's ::MY
Isn't it?
luqui no, it's MY::
Limbic_Region schock - p6 should mostly still let you do that
Juerd luqui: That's ::MY:: :)
luqui but %MY is a namespace hash
integral PerlJam: (NEGIB { $*compiler}).get_current_subrtouine ;-)
Juerd luqui: :: is the namespace sigil too
luqui (maybe it's %MY::) 15:07
Juerd luqui: Then that can be %MY or %*MY
luqui Juerd: er, symbol table
Juerd But I really, really don't care where it comes from.
luqui %*MY doesn't make sense, because it's not global
schoch i never liked the BEGIN{} either though
luqui schoch: what would you prefer?
PerlJam schoch: why not? I've always thought BEGIN was one of the things perl did right.
Juerd luqui: Why the heck care? In Perl, scope isn't visible in variable names. Type is.
schoch not sure but it behaved oddly with some modules
PerlJam schoch: Though my eyes tend to glaze over when you talk about INIT and CHECK subs ;) 15:08
luqui Juerd: and scope. $^ is a scoper. $* is a scoper (but you're arguing it)
$. is a scoper
Juerd luqui: Declaration defines scope, not the sigil. In Ruby, sigils are for scope, in Perl, they're not. This shouldn't be any different for special variables.
PerlJam schoch: BEGIN has never behaved oddly. It's always quite predictable
Juerd $.foo is just a lexical declaration
luqui Juerd: then what are secondary sigils for, if not scope?
schoch not with par
Juerd It happens to be an alias for the foo attribute of the current invocant
schoch i don't rememeber the details...
Juerd But that's not a scoping issue. 15:09
Alias_ I like $.foo
schoch goes back to lerkin
Juerd I like it too.
luqui Juerd: so what do secondary sigils stand for?
Alias_ It's special in a predictable way
luqui Juerd: what does even one secondary sigil stand for?
Juerd luqui: When things are defined, where they come from, and certainly not scope...
luqui Juerd: example 15:10
Juerd $*PID, * indicating that it comes from the system, as with %*ENV
luqui to me, the secondary sigils represent those scopes that aren't declared.
Juerd $?OS, indicating it came from the compiler
&?SUB, etc
luqui the * sigil means simply "global"
nothing more
Juerd Exactly how global is %*ENV when you've made it local, which should be possible in Perl 6? 15:11
PerlJam luqui: well, * is runtime global, and ? is compile-time global n'est pas?
autrijus temp %*ENV
luqui Juerd: you know how temporization works better than to ask that, I think
Juerd Eh, ignore that.
luqui PerlJam: ? is compile-time lexical
Juerd luqui: Yes, a stupid question it was.
PerlJam luqui: er, right.
Juerd luqui: Why are we indicating scope? 15:12
luqui because we don't want to declare $^a
Juerd And why can't secondary sigils just mean they're *special*?
luqui well, they could
Juerd And have only three
* for special
luqui but so far all the specials indicate types of scope
Juerd . for aliased to public attribute
: for aliased to private attribute
luqui (sometimes odd ones, like $.)
PerlJam Juerd: and ^ ? 15:13
Juerd OR be consistent, and remove our/my
And use secondary sigils for that.
luqui Juerd: and state
Juerd PerlJam: I don't know that one. What does it currently do?
luqui: Yes, and state.
luqui sort { $^a <=> $^b }
Juerd Oh, right
PerlJam Juerd: implicit sub parameter
Juerd That's a lexical thing
luqui it's implicitly declared
Juerd They're special in that they're implicitly declared and they're actually parameters too 15:14
luqui Juerd: well, I don't know the deeper reasons...
Juerd Like with $. and $:, they're aliased to other values.
luqui I've given a good argument, I think, but you might just have to ask larry for this one :-)
I know I like it
Juerd The problem with ., : and ^ is that people think they indicate scope.
They do not - the scope is lexical 15:15
However, the variables are bound/aliased
luqui well, . isn't exactly lexical
Juerd It is too.
$.foo is one alias for many variables.
luqui hmmm... I guess since method Foo::bar () is outlawed
You could think of it that way 15:16
Juerd But the scope of the variable is defined entirely outside the class
luqui Or you could think of it as "object scope", which is a weird kind of dynamic scope
Juerd The scope is determined by construction and destruction
Not by anything else
Just like how our creates a LEXICAL alias for a GLOBAL variable 15:17
jabbot pugs - 1278 - * Implement q/.../ string literals
Juerd The scope is determined by that the variable is never destroyed except at the end
luqui Juerd: ok. there are a lot of ways to think about abstract programming language concepts.
Juerd Not by where the variable is declared.
It's really important to separate variables from their names 15:18
$foo is not a variable, it's just its name
luqui that depends on how you're defining variable
Juerd Well, explain to me what $foo := $bar does if $foo is anything other than a name.
rgs woot.
Juerd It binds the name $foo to the variable that is also called $bar
It's the same variable, though
Variables don't exist at syntax level 15:19
We call these names "variables" in every day English, because that makes discussing things a whole lot easier
luqui take: $foo is a container, as is $bar, for a thing that holds values. := takes the thing that holds values in $bar and puts it in $foo.
Juerd No, container == variable 15:20
Two containers cannot both contain the same value
(Note: same isn't the same as equal)
luqui Juerd: not always. you're getting into semantics here.
Juerd Think of real life containers
PerlJam Juerd: sure they can! Superpositional containers :-)
Juerd You cannot put the same apple in both
luqui And I think you're beginning to define variable differently from what we think when we say "varaible"
and perl is designed to coincide with how we think
Juerd luqui: When we say variable, we usually mean mutable container.
schoch how do references work in this context? 15:21
Juerd schoch: A reference is a VALUE (that lives in a single container, that in turn can have any number of names) that refers to any single container
Possibly its own container.
(Circular-reference-in-one)
Names, containers and values are the three things a programmer/programming language deals with 15:22
luqui Juerd: think about what my proposals do. They take the existing language, rip out its guts and replace them with new ways to think about it without changing the syntax or semantics (or if so, in a very slight way).
Juerd Names exist at syntax level
luqui There's more than one way to think about thing.
s
Juerd They're how we, programmers, address variables
Containers are locations in memory. It's how the computer accesses the values
luqui and language design is about finding the right way to think about things
and containers are certainly more than locations in memory for a high level language 15:23
Juerd And the values are the actual values, and cannot be addressed by anything other than their container. A single scalar container has one value (disregarding dualvars), array or hash containers have many values, they being other containers)
luqui In other words, you're cramming a way of thinking down our throats (not to say that that's bad, there's a famous saying about that :-)
When we're not even sure whether the way we're thinking is right. 15:24
Juerd Given how references and objects work in Perl, this is a good way of thinking
luqui And Larry is _incredibly_ good at thinking about _everything_ in many different ways.
Juerd This model can't easily be used with, for example, Python
Where "same" and "equal" are the same thing.
luqui So, try and think how larry is thinking when he introduced $? before saying that it's the wrong way.
Juerd And values are immutable.
luqui: I haven't said it's wrong. 15:25
luqui okay, s/saying that it's the wrong way/when disputing it/
Juerd luqui: I think it's unnecessary to add another sigil
luqui so it's an aesthetic issue?
Juerd Certainly.
luqui because you could have said that up front :-)
and then the discussion is much less deep and much easier to, er, think about
Juerd You don't have to be able to see what kind of variable a variable is 15:26
And if we're convinced that it is important, then we should ditch declarators like my and our, and replace that with sigils too.
I don't understand, for example, why has is spelled has, if there's ALSO a secondary sigil
luqui and then what does the null sigil stand for?
Juerd luqui: Null sigil stands for subroutine call or namespace :) 15:27
luqui Juerd: because larry "got tired of reading C++ code and wondering whether something was a member or a global"
jabbot pugs - 1279 - Remove dead code corresponding to the ol
luqui Juerd: sry, null secondary sigil
null twigil
Juerd luqui: I know people who got tired of reading Perl code and wondering whether something was lexical or global. They have found a solution in Ruby.
autrijus trigil! quaril!
Juerd luqui: User defined non-special variables (that is: names for them), I hope 15:28
Possibly user defin-able.
luqui like lexicals?
Juerd Scope is irrelevant.
luqui so then what does my do?
Juerd my initializes a container and declares a name for it.
luqui it defines one of these user defined non-special variables
how about that
Juerd The name is lexical
The container is destroyed when there are no more names (references) left 15:29
(Either at block end or a more convenient moment)
But the container CAN live on long after the name is no longer attached to it
luqui so um, what does the null twigil, in the new system where we've ditched my and our, mean?
Juerd I wouldn't want to think of that system, because I like my and our, and not having twigils/sigils for scope. 15:30
luqui i'm trying to get you to think about something
Juerd But it could indeed mean the variable is lexical. That's huffmanly correct, and the way Ruby does it.
luqui and it does
currently, even
Juerd Huh? 15:31
luqui which lexical scope it is attached to is defined by "my" or "our", though
Juerd No, a user defined (twigilless) name is global unless declared otherwise
no strict;
$foo = 1; # not lexical!
autrijus use autolexical;
$foo = 1; # lexical!
autrijus grins
Juerd Yes, but that isn't default behaviour
luqui Juerd: I'll say treat that as a special case
autrijus Juerd: yours isn't the default.
the default it neither.
luqui the default is "$foo not declared" 15:32
autrijus yup
Juerd luqui: The scope of every explicitly declared name is lexical.
PerlJam autrijus: not in my one-liner it's not.
Juerd our, has, my
The declaration, and thus the name, is lexical
luqui PerlJam: yeah, that's the "no strict" or "use autoglobal" case
autrijus PerlJam: maybe oneliners uses autolexical :D
j/k.
Juerd The variable itself (the *container*), is global, object-dynamic, or lexical
autrijus: It depends on whether you see strict as the default, or no strict as something you can load. 15:33
luqui Juerd: exactly
and both ways of thinking define, conflicting, ways to think about what a twigil means
Juerd Yes, but now both ways of thinking are in Perl at the same time 15:34
luqui yep
which is what perl is best at
Juerd We use twigils to indicate scope, but only for special variables, because for normal user-definable variables, we use declarative keywords.
That's inconsistent in a way that IMO gains nothing
It's not as if the distinction between $* and $? makes it any EASIER to use 15:35
PerlJam btw, perl6 was starting to solidify ever so slightly in my head ... until I started paying attention to you guys ;)
luqui other than the fact that we've defined a meaning for the null twigil
Juerd To the contrary: you have to know more.
luqui PerlJam: solution, don't listen. We're rambling theorists :-)
PerlJam luqui: sure, but I had a belief of what was "right" and you've both shaken my belief 15:36
luqui Juerd: $* is global. you can't say $*SUB because that's not global.
rgs autrijus: my three last commits are heavily cargo-culted. beware.
Juerd luqui: I'd love to ramble some more, but I can't because my hands hurt beyond what pain killers can fix
So I will have to stop now.
Sorry.
luqui do you have carpel tunnel or something?
Juerd Not exactly that, but similar
autrijus luqui: yes. (to save Juerd's typing)
luqui hehe
any suggestions about how to avoid that? (I use my hands for everything) 15:37
supposing you can type it
Juerd luqui: www.kinesis-ergo.com
luqui (don't let me impose)
jabbot pugs - 1280 - Factorize code between qLiteral and qqLi
nothingmuch Juerd: get Jouke to make a vocal IRC for you, or something ;-)
Juerd luqui: www.dvorak.nl
luqui okay
Juerd luqui: www.wacom.com
luqui: Combine these three and you'll postpone it a long time
afk
nothingmuch these guys are pretty cheap: store.ergocube.com/
because they are pretty much nonprofit 15:38
luqui cheap enough not to have a website
n/m typo
autrijus journaling, bbiab
Juerd nothingmuch: They sell a lot that has nothing to do with ergonomics, or makes things worse. They also have the good stuff.
afk
autrijus rgs: if it passes tests, cargo code is just fine :)
nothingmuch luqui: also excercise a lot while you can 15:39
autrijus rgs: and it looks ok.
luqui nothingmuch: uh oh
nothingmuch Juerd: i have not yet bought any ergo stuff
rgs ok, then. bbl &
nothingmuch luqui: stringed instruments have kept my writsts in excellent condition
running in highschool ruined my knees
luqui plays piano and violin
nothingmuch guitars & double bass 15:40
Juerd luqui: Ouch!
luqui where "plays" is defined loosely for violin :-)
PerlJam luqui: ouch. you've tripled your chances for RSI at least
Juerd luqui: 1. violin 2. computer 3. piano - top three rsi causes!
afk
nothingmuch but they must be played properly, ofcourse =)
Juerd (for real)
luqui I know. that's why I'm asking for suggestions.
nothingmuch piano - i don't know about
PerlJam luqui: rest your hands.
nothingmuch violin - i've heard that it is stressful to hold it like that 15:41
luqui it is... which is one of the reasons I don't practice much
it hurts to do it for very long
nothingmuch on the bass everything is very round, and you are never supposed to do anything stressful. If there is discomfort, then it will sound horrible, and also hurt you
and if it won't sound horrible, it will limit your progress
the thumbing position for example, is very dangerous in this respect 15:42
PerlJam luqui: For the computer and piano you can use alternate forms of input (chorded keyboards with big fat keys, foot pedals, etc). You're pretty much stuck with the violin though.
nothingmuch i've been having lots of trouble with it
luqui I imagine the same for violin.... at first piano was stressful, but now I think it relaxes my hands
nothingmuch but my teacher never ever ever lets me go the cheap way
i could just press harder, but that would ruin my thumb in no time, especially with my mother's genes controlling my joints 15:43
PerlJam luqui: you could always redesign the violin. Make an instrument with the same tonal qualities but less hand pain :)
luqui PerlJam: I'd love that :-)
play it "topwise"
Juerd re 15:44
nothingmuch PerlJam: that's what they've been doing for a long while, y'know ;-)
was it paganini who ruined everything by making people stick it in their neck?
luqui when was paganini?
Juerd luqui: For any keyboard, make sure its keys are just heavy enough for your hands, and no key feels much different than the others. Make sure the feedback is good, both in feeling and hearing
PerlJam nothingmuch: "they" perhaps aren't as smart or creative as luqui (or other perl people for that matter ;)
nothingmuch i'm not sure, really
luqui 1782-1840 15:45
Juerd luqui: Church organists get rsi more quickly because of the delay there is between hitting the key and hearing the sound.
luqui I know beethoven stuck it in his neck
Juerd That's really bad feedback.
luqui huh, why does feedback make a difference?
nothingmuch i wonder if oboe players have rsi in their mouth ;-)
luqui lol 15:46
metaperl lol
nothingmuch luqui: you automatically do more than you need
luqui used to play oboe... yes.
Juerd The better the feedback is, the more you know (subconsciously) what the right pressure is, so you don't use more force than necessary
PerlJam The only time I ever played violin it felt completly natural ... Except that I can't play, so it probably sounded completely horrible too :)
rgs plays piano, trumpet and trombone
Juerd luqui: Avoid cherry keyboards (silent and "soft touch" feeling) at all cost.
PerlJam Juerd: those people need more faith!
luqui oboe is stressful on the mouth.
nothingmuch wants to learn trumpet
luqui Juerd: yay, I hate those things!
metaperl I want to learn guitar/sitar
nothingmuch luqui: what about the delay? how big is it? to me it's legend more than anything
metaperl: guitars are probably the most accessible instrument after piano 15:47
Juerd luqui: For this reason, I don't think touchstream keyboards are any good. I'd have to try one first to be sure, though.
PerlJam metaperl: me too, but my fingers are too fat. (also why the violin sounded horrible I'm sure)
nothingmuch technically and otherwise
price, popularity, etc
luqui and even moreso than piano if you consider price
nothingmuch PerlJam: most people fingers aren't too fat for the guitar
luqui nothingmuch: delay.. for oboe?
nothingmuch that's what my dad told me 15:48
oops, ride is leaving
luqui hadn't played any instruments before, so I didn't really have anything to compare it to
and didn't get good enough
PerlJam nothingmuch: well, I think I'd need a wider neck on the guitar but then my fingers wouldn't be long enough
metaperl PerlJam, never underestimate yourself
Juerd I want 6 fingers on each hand.
metaperl Django Reinhardt kept playing after only having 2 fingers
luqui playing Liszt will stretch your hands as much as you need :-) 15:49
Juerd Or a shorter alphabet.
luqui so I imagine your fingers could grow to a bigger guitar
Juerd: just six?
Juerd Yes - with more than six, the keyboard would really require redesign 15:50
luqui I want eight, because then I could count in hexadecimal
Juerd And we all know how long it takes for people to create a sane keyboard.
luqui haha
Juerd Why count in hex if you can already count to 31 on a single hand, in binary?
31 is almost twice as much as 16 :)
luqui you could count to 256 then 15:51
Juerd Yep.
luqui (255)
Juerd +(@foo = 0..255)
[0..255].elems
luqui is possibly one of the fast finger-binary counters ever... it just takes him a while to read back what he's counted to
Juerd Probably +[0..255]
metaperl playing piano was great for my career in the video game Joust 15:52
luqui Joust?
rgs trumpet trains you to binary-count with 3 fingers.
metaperl you needed fast fingers to get the bird to flutter up
Juerd We used binary in school to communicate test answers
metaperl yes, a video game from the 80s
Juerd 31 > 26, so enough for the alphabet
And after 3 letters you have enough to fill in the rest
And it really draws NO attention if you keep your fingers horizontal 15:53
luqui except for when you stop at 4 ;-) 15:54
Juerd Yes, the D can be troublesome.
luqui eh, I never liked 4 anyway. It was the first of the dreaded composites.
Juerd We had an IRC channel we used with several people from my class
On it, "132" meant a certain two fingers :) 15:55
So we used 4 and 132, in decimal (shorter)
Even the non-nerds copied that. 15:56
Hm, alias didn't like it... :| 15:57
jabbot pugs - 1281 - more Pod::Stream::Parser; improvement of
luqui Off to school. bye 16:00
luqui &
obra a 16:30
Juerd I hate my brain 16:31
My brain has decided that bitwise ops should be ?+ for OR, ?- for XOR and ?* for AND
And that stringy and numeric should thus be ~?+, ~?-, ~?* 16:32
And +?+, +?-, +?*
It's a good thing I'm not inventing this language
mugwump *hic* 16:45
jabbot pugs - 1283 - Move PhD logo to advocacy directory 16:47
pugs - 1282 - First part - the DOM tree structure
Juerd tnx.nl/3689VBOF # mad consistency. 16:49
gaal integral, still aroung? 16:55
integral gaal: yep
gaal i was afk and backlogging now i noticed oyu wanted to workon %CONFIG 16:56
didn't finish reading up yet, but before you do anything,
there's a thread where larry explains his plans for it
doesn't look decided
~6 March on p6l i think 16:57
or search these logs for config
integral oh, ok. I asked autrijus and he just said go ahead. I'm just exposing the FM in Config.hs as a hash
gaal just wanted to warn you so you don't work a lot for nothing - sorry if someone already
ESCdd
okay :)
but be aware that this is specifically likely to change... 16:58
i had these same toughts myself.
integral oh, I've no problem with my code being ripped out :-) It's to access some pugs specific conf data though, so it's a hack anyway
elmex hi
gaal my %conf = map { split /:/, $_, 2 } `pugs -V`; # j/k :) 16:59
Juerd chomp. 17:13
gaal true, but i was being silly.
Juerd Then so am I :)
gaal also, the current separator is ": "
and there's a leading \t 17:14
and there's a header line
and there are some @INC entries too 17:15
Juerd is going to try to take a long break
As you can see on p6l, I'm having too many thoughts right now :)
gaal yeah, and outsilliying contests are tiring :)
Juerd hehe
Now you're just being silly
:)
gaal bows 17:16
Juerd I'm quitting - you win :P
(this time)
gaal what are "foes" on use.perl.org? 17:17
lumi Like in /. I think, the opposite of friends 17:18
And freaks are the opposite of fans
gaal i don't do /.
i have a feeling that outsillies me
so like if friends are those who read you regularly, foes are people who regularly don't read you? 17:19
crysflame freaks are people who are friends of your foes, i think
lumi No, friends are people whose opinion you value
And foes are the opposite 17:20
Freaks are people whose foe you are
crysflame aha
lumi (That parses, honest!)
crysflame freak, foe, fan, fum
gaal and that information needs site suppose because...?
lumi Because you could give the people you value a bonus, so you'll see their writeups more
And people who value your opinion could give the people whose opinion YOU value a bonus 17:21
That's the theory, I think
e2 makes more sense though
Or not
gaal i like my feeds a little more deterministic than that :) 17:22
lumi It's for comments, not for main entries, as a rule
gaal anyway, i didn't mean to go off topic for so long, lets take this off-channel
elmex hi 17:40
gaal ho
elmex hmpf 17:47
how fast is pugs at the moment?
cm schnell genug! 17:48
elmex lol
cm: _how_ fast 17:49
cm schnell wie der wind, wie der wihihind..
elmex: 370
elmex in comparsion to eg. perl5
cm i've no idea, i'll stop trolling :p
elmex cm: cuteee ;)
cm =( 17:50
PerlJam elmex: It is about as fast 17:52
(as perl5)
elmex PerlJam: with pugs compilin to parrot? or interpretion? 17:53
PerlJam both. Slightly faster when compiled to parrot, slightly (a little more so though) interpretted
But all in all, they are about the same speed 17:54
elmex hmm, wonder how fast it may be with a nicely jit'ed parrot 17:55
cm if it chips like a parrot, it surely is a parrot ;o 17:56
elmex ? 17:57
cm: cuteee ;) are you female?
cm o_0 17:58
xerox ōæ½xC2
Whoops, sorry.
elmex cm: c'mon 17:59
cm elmex: hm? 18:01
xerox elmax /whois cm 18:02
elmex xerox: no sign for male gender 18:04
xerox elmex: 1) who cares 2) the name isn't enough? 18:08
elmex xerox: ;) 18:12
kungfuftr lo all 18:15
Corion BTW - NickServ pesters me - should I really register my nick ? 18:19
elmex if you want to, yes ;)
Steve_p freenode.net/faq.shtml#nicksetup 18:21
stevan autrijus: are you still awake? 18:25
kungfuftr shit... another tsunami 18:27
Jouke yes? sure? Last I heard on the news was that there might be a new tsunami, but that it wasn't sure yet 18:31
integral see ##quake
kungfuftr news.bbc.co.uk has a live video feed 18:32
Jouke reads about 'tsunami warnings' on CNN and bbc.co.uk, but not an official tsunami report yet 18:35
jabbot pugs - 1284 - Shifting argument to subroutine fails 18:37
gaal Corion, you there? 18:42
Limbic_Region Corion is jaw jacking at PerlMonks about his bike - you want me to get his attention?
Corion I'm also here :) 18:43
gaal hey :) just a note about the parsefails since i read logs from last night :
Corion (and I get an annoying beep whenever somebody mentions my name here :) I should switch that off)
Limbic_Region how's that for timing ;-)
gaal nm and i are at no disagreement.
if you get away with todo_is_eval on a test that doesn't parse, great. 18:44
Corion gaal: Ah! I think I got away with that, yes :)
gaal if the parsefail is "hard", though, it's b0rk eval, even, and that's what you need the comment + warning note in the form of a failing test for.
Corion Win2k:r1284 (118/3242) 18:45
gaal but since this is a kludge, we agree on a convention for marking these things.
Corion ... so we now have scripts that fail their tests again :)
gaal fails are ok
Corion gaal: Aaah. OK - I hope I stay away from the eval-borking tests :)
gaal no, evalb0rktitude is fun :)
Corion gaal: But somebody converted lots of fails to todo_ stuff lately :)
gaal i, however, have a chance to go home now, and therefore shall now take my leave of you, kind sirs. 18:46
Corion Oooo - somebody must have fixed the warning that my code created yesterday - I'll have to look at how, so I learn :)
Corion is already home :)
gaal & svn blame is your friend
kcwu i fixed it, but i don't know why it fixed ;) 18:48
Corion kcwu: Heh :-) 18:49
kcwu: Ah - but that's exactly the fix I imagined - you have to tell Haskell somewhere what types you expect it to use for the numbers
kcwu: And you introduced Int as the type to use, so Haskell is happy now :)
... now, BEGIN {...} blocks need to be made to work, I guess :) 18:50
kcwu why tell haskell that? i image haskell will deduce that it is number, and use (compare) them without problems 18:53
Corion Does anyone know off-hand how Perl5 handles C<-l> ? Is it translated into Perl code, like C<-e> ? Especially if you use "perl -l script.pl" ...
(if nobody knows, I'll go source diving :) )
kcwu Corion: i am sure is it you want: perl -MO=Deparse -le 'print 1' 18:54
chip Anybody remember whether Perl 6 /foo/ is more like Perl5's /foo/ or qr/foo/? 18:55
autrijus chip: yes.
Corion kcwu: Yeah - I know what Deparse outputs, but I don't know if that is some special fudging of B::Deparse or if Perl really created that code :-)
autrijus (go not to autrijus for counsel...)
chip I can't remember whether e.g. C<$a of Str where { /^a/ }> is good Perl6 or not
autrijus chip: in value context it does the match.
otherwise it is rx//. 18:56
kungfuftr hhhmmm... is there a simple way in perl6 of declaring many constants at once?
chip "value context"? That's not a context I know
autrijus chip: i.e. not in direct RHS of a binding or assignment.
Corion kcwu: But my plan is (currently) to translate C<-l> to C<-e "BEGIN{...}"> - but then I get to the problem of having both, a file AND a -e fragment. I could move the whole file-loading into Perl, of course :-))
autrijus chip: it's specced in the synopsis, although I question its sanity a bit. 18:57
chip I guess my cunning plan to avoid scanning the [AES] docs has been thwarted
Corion svn blame is really cool btw ! I should add something similar to my Javascript SubEthaEdit clone :-) 18:58
autrijus Corion: wait, your *what*!?
Corion: url! url! 18:59
Corion autrijus: zaphod.datenzoo.de - very early beta
FireFox / Mozilla preferred, but it works in IE6 as well. Opera is dodgy
autrijus holy zaphod. 19:00
I can now sleep with the belief that there is hope in this world. :)
autrijus waves &
Corion autrijus: VERY early beta
Corion gave autrijus JS nightmares
Corion watches the children at play 19:04
(but some times, the children just time out :) ) 19:05
kungfuftr pugscode.org <<Overview Journal>> | pugs.kwiki.org | logged: xrl.us/e98m | FBSD_4.11:r1284(118/3242) win2k:r1284(118/3242) Linux:r1197(105/3116) MacOSX:r1197(105/3116) 19:06
Corion realizes just now that anybody can change the channel topic... 19:07
so I don't have to report my Win2k test results but can update the topic myself! D'oh ! 19:08
kungfuftr yup
there's some sort of DBI lite module somewhere, right? 19:26
Corion kungfuftr: Is there? There is DBD::SQLite, but I don't know of any Perl6 thing dbi-ish 19:27
kungfuftr Corion: nah, i meant a lightweight version of DBI 19:28
kungfuftr has been looking at porting DBI::PurePerl, but it's a git
Corion kungfuftr: Not that I'm aware of - DBI is the only DB interface that Perl has (although I use `isql` :) )
kungfuftr: I was pondering to learn Haskell to implement some smallish SQL machine in it (it would rock to have a built-in SQL DB for Perl6 :) ) 19:29
kungfuftr: But you could look at porting SQLite, together with DBI :-)
kungfuftr SQLite might be possible, but DBI is very very complex and uses a lot of perl5 specific things 19:30
Corion kungfuftr: Ah - I would rather consider a reimplementation of the API, or at least connect, prepare, execute and fetch :-) 19:31
kungfuftr DBI::PurePerl, is really DBI in disguise... it is used via an environment variable 19:32
Corion kungfuftr: But I haven't looked at anything of the internals. Writing an SQL DB in Haskell should be "pretty easy", as Haskell can do parsers and lazy lists. It wouldn't be very efficient/performant, but bad performance is purely a hardware problem
kungfuftr ie: $ENV{PUREPERL} = 2; use DBI; # will use DBI::PurePerl instead of the XS based dbi
i reckon porting of some more core modules like Symbol.pm etc would be required 19:34
gaal corion, you still at -l? 19:38
Corion gaal: No - currently I'm hacking on a test to verify (multiple) -I switches (uppercase i, not lowercase L)
gaal perl.c:2756, more or less. :)
Corion gaal: But I plan moving back to -l tomorrow :) Oh - thanks for searching the stuff ! :)
gaal search? what search? :p 19:39
Corion gaal++ # knows perl.c by rote
Anyone got a good idea how to print out @INC (newline or \0 separated) without mentioning $_ ? pugs -e "print for @INC" doesn't compile for me 19:40
(also, other shell metacharacters should be avoided, of course :) )
gaal by heart, you mean -- and actually no, it just happened to be open in some screen window in the past few days :)
@INC.perl.say
Corion gaal: Yes - I meant "By Heart" :)
gaal: Ooooo ! Cool (except that I now have to parse that a bit more, but that's OK :) ) 19:41
gaal ./pugs -e 'map {say $_}, @INC' 19:42
Corion gaal: That mentions $_
gaal ah
so it does :)
./pugs -e 'map -> $x {say $x}, @INC'
Corion gaal: I want to stay as platform agnostic as possible, that means that the code should run in double quotes (Win32) and should not mention $_ (UNIX shells)
gaal see latest 19:43
Corion gaal: It uses $x - which is as bad as $_ :-) No $ sign in shell oneliners
gaal aha!
Corion gaal: I'll go with the @INC.perl.say - documenting/introducing Perl6ish oneliners is also important!
kungfuftr hhhmmm... any idea about how we'll be doing tie() sort of stuff in perl6? 19:45
Corion kungfuftr: Not at all, I hope ! :-) 19:46
Khisanth isn't that what the my $foo is Bar; type of stuff is for?
kungfuftr yup, i would assume so, since hashes are just objects are anyways
gaal see e.g. <L09/"PDL support"> 19:47
L<S09> that is
kungfuftr Khisanth: yeah `my %hash is ixHash;` is what i'd reckon
gaal interesting short circuit, that.
Corion Grrr - the PCRE implementation seems to match dodgy ( or ~~ is broken ... ) - do we have C<like> implemented in Test.pm ? 19:48
kungfuftr has a read through the synopses again
Khisanth hmm is there an 'of' in perl6? :)
Corion Because "ok $got ~~ $expected" doesn't DWIM, even though $expected = rx:perl5/^('foo/; :-(( 19:49
(it's always true)
gaal Corion, we don't have like yet. 19:50
Corion gaal: Ah - most likely, ~~ not DWIMming is the reason :-) So I'll go with substr() instead :)
(except that now, the newline stuff becomes uglier...)
gaal then embed perl5 code. }:) 19:51
Corion gaal: :-) 19:52
lumi What I wouldn't give for Perl 6 rules, today
In the next 10 minutes, in fact
stevan Corion: regexp support is spotty 19:53
best not to do like() yet
Corion stevan: Looks like it :)
Anybody got a better idea than substr() to check if the first/last two lines in a scalar match what I expect? 19:54
stevan most of what works, autrijus put in because I begged him for it :)
gaal Corion, are you compiling a set of known p6 one-liners? 19:55
that would be a good node for perlmonks :)
grep around for eyespoplikeamosquito's first ever p6 japh, which i don't understand 19:56
Corion gaal: Nope - I'm just writing a test script that checks that -I works properly. But your @INC.perl.say would be a good "verification oneliner" (that is, it should go into the docs IMO) - it has no shell metacharacters.
gaal in that case, ./pugs -I /idontexist -V | grep idontexist should work too 19:57
(does in p5)
ninereasons it's going to eventually be "print for @INC"; but that won't work right now, right?
gaal ah, also works in pugs! :)
yes 19:58
Corion ninereasons: Yes. And also, it will require the user to put the stuff in quotes, which is error prone. That's what I like with @INC.perl.say - one shell-word
gaal: I don't like grepping - that's error prone. I'm especially interested in the _order_ the stuff goes into @INC
rgs what is @*INC.perl ? 19:59
gaal Schwern was here a few days back saying that @ can be a metacharacter in some makes, ifyou care about that
elmex f
Corion But I think I know how to fudge it - I simply "initialize" my reference scalar from a single unadorned call to pugs.
gaal rgs: think "Data::Dumper::Dump"
elmex ?
Corion gaal: Yes, but I don't care currently. This oneliner would be for shell verification and my test file :-)
rgs wants a @*INC.yaml.say 20:00
gaal rgs, you're sure to get it when YAML.pm lands
in fact grep the channel logs for autrijus' explanation of how easy this kind of thing is to add 20:01
(modulo YAML.pm itself :-)
kungfuftr needs to buy the perl6 book 20:02
rgs gaal: is it planned ?
gaal what, a yaml module for p6? of course! if for some reason ingy doesn't port it i'm sure someone else will 20:03
Limbic_Region has autrijus mentioned what the plan is on p6 rules/regexes is? 20:05
nevermind
# 6.283: Rules and Grammars.
xerox Limbic_Region: url?
Limbic_Region svn.perl.org/perl6/pugs/trunk/docs/...rview.html
Corion Err - if I have an array reference $t = [ "foo", "bar" ], what is the Perl6 equivalent to Perl5 @$t ? 20:06
Limbic_Region xerox, if you haven't seen www.pugscode.org you might want to check that out too
as well as autrijus's use.perl journal
use.perl.org/~autrijus/journal
xerox Danke. 20:07
ninereasons $t.elems 20:08
I think, Corion
Corion ninereasons: Looks good, at least no compiler error there anymore :) 20:09
ninereasons I'm wrong, though. That's not a deref, as you asked for. 20:10
Corion ninereasons: Yeah - I just fixed the other bugs, and now Pugs tells me that it doesn't know about .elems :) 20:11
ninereasons a TODO
Corion Anyway - I get a real life timeout now, so -I test-hacking will have to be postponed. 20:13
later all!
gaal so, is it Boolean (S04) or Bool (S06)? latter wins, right? 20:30
schoch hmmm 20:33
schoch is reading synopsis
junctive types seems scary 20:34
gaal also: how are constants spelled? is thre no sigiless form? my $val is constant = 8;
how does a constant imported from antoher class look like?
kungfuftr yay! looks like i can finally signup for safari! 20:39
kungfuftr pokes activestate people about getting special rates
Khisanth why poke activestate?
wouldn't you want to poke O'Reilly? :)
kungfuftr Khisanth: activestate have a special safari arrangement... my company owns activestate 20:40
=0)
Khisanth hmm
kungfuftr Khisanth: ActivePerl Studio Perl includes a safari subscription 20:41
Khisanth only uses activeperl occassionally :) 20:42
kungfuftr uses it only when he has to, but he gets it for free 20:44
Khisanth well I only use it on windows and with the lack of a decent shell/terminal it isn't very fun 20:45
kungfuftr Khisanth: some of the pro dev tools help a lot in that respect 20:47
Khisanth assumes a bunch of GUI stuff
kungfuftr some commandline stuff too 20:50
schoch wmic is quite nice 20:55
21:00 brentdax_ is now known as brentdax
kungfuftr wmic? 21:05
schoch wmi command line 21:06
can do some crazy shit with it
jabbot pugs - 1286 - Regression test for -I 21:07
pugs - 1285 - Fixed typo
schoch kernel panics are fun 21:08
kungfuftr for example you can do class win32_processor get MaxClockSpeed
on a remote box
21:09 hide_ is now known as hide
kungfuftr ah 21:09
wilx Hmm, what good is it for? I've never figured out what can onde do with WMC.
s/onde/one/ 21:10
schoch WMI you mean?
wilx Err, right. WMI.
wmic is the executable, right?
schoch lets say you need to figure out every winblows box on your network that has an admin user
it makes that very easy
wilx Hm.
schoch almost everything in windows has a WMI interface 21:11
21:11 metaperl_ is now known as metaperl
kungfuftr schoch: search.cpan.org/~wyant/Win32-Proces...nfo/WMI.pm ? 21:12
schoch i used it once to coordinate remotly start/stopping perfmon for a stress test accross several web boxen
wilx Heh, nice.
schoch www.windowsitpro.com/Windows/Articl...23854.html 21:14
i'm sure it can be used for dastardly evil stuff too 21:15
jabbot pugs - 1288 - renamed test for -I option 21:17
pugs - 1287 - 'make test' ignores test files with uppe
kungfuftr hhhmmm... does `%hash.kv -> $k, $v` bind variables? 21:45
stevan kungfuftr: it should (i think), but i dont know if pugs can handle it yet 21:46
jabbot pugs - 1289 - adding some io-in-loops tests 21:47
kungfuftr stevan: hhhmmm... tbh, i can't really see a good reason to do it by copy as posed to bind
stevan oh 21:54
hmmm, didnt read the bind part of the question
I would expect the keys to be immutable, since to do it any other way might get a little insane 21:55
then I would also expect the values to do that as well 21:56
schoch krazy keys
stevan and use %hash{$k} = $new_value if you wanted to change something
kungfuftr stevan: yar, i suppose, not very dwim though
stevan I recall reading something somewhere about hash interators and thier caveats and limitations
kungfuftr: what does p5 do with each(%hash) 21:57
I think it copies IIRC
schoch iterators in java have some major problems in multithreaded apps
stevan iterators in any langauges can have problems multithreading :)
kungfuftr yar
schoch ya, the locking can be painfull 21:58
stevan (except for some weird parrallel logic/dataflow languages like Lucid)
schoch will perl6 have any concept of threadsafe?
stevan schoch: I have no idea,.. my guess is that they might differ that to Parrot 21:59
or at least as much as the can possibly differ to parrot
ingy: ping? 22:02
ingy: just wondering if your battle with MakeMaker re: Test.pm and File::Spec was fruitful 22:03
jabbot pugs - 1291 - added END block to io tests to remove th 22:17
pugs - 1290 - more Pod::Stream::Parser work; better ve
chip Larry's last statement on threading was that Perl would expose whatever Parrot provided 22:24
At least, that was his last statement in my personal reference frame
the grass is always greener on the other side of the light cone 22:25
gcom 22:37
kungfuftr (p6 and parrot essentials)++ # good reference 23:27
ingy stevan: hola 23:54

Powered by Cro and the RakuĀ® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!